Request for Comments: 4105 France Telecom
Category: Informational J.-P. Vasseur, Ed.
Cisco Systems, Inc.
J. Boyle, Ed.
PDNETs
June 2005
Requirements for Inter-Area MPLS Traffic Engineering
Status of This Memo
This memo provides information for the Internet community. It does
not specify an Internet standard of any kind. Distribution of this
memo is unlimited.
Copyright Notice
Copyright (C) The Internet Society (2005).
Abstract
This document lists a detailed set of functional requirements for the
support of inter-area MPLS Traffic Engineering (inter-area MPLS TE).
It is intended that solutions that specify procedures and protocol
extensions for inter-area MPLS TE satisfy these requirements.
Table of Contents
1. Introduction ....................................................2
2. Conventions Used in This Document ...............................3
3. Terminology .....................................................3
4. Current Intra-Area Uses of MPLS Traffic Engineering .............4
4.1. Intra-Area MPLS Traffic Engineering Architecture ...........4
4.2. Intra-Area MPLS Traffic Engineering Applications ...........4
4.2.1. Intra-Area Resource Optimization ....................4
4.2.2. Intra-Area QoS Guarantees ...........................5
4.2.3. Fast Recovery within an IGP Area ....................5
4.3. Intra-Area MPLS TE and Routing .............................6
5. Problem Statement, Requirements, and Objectives of Inter-Area ...6
5.1. Inter-Area Traffic Engineering Problem Statement ...........6
5.2. Overview of Requirements for Inter-Area MPLS TE ............7
5.3. Key Objectives for an Inter-Area MPLS-TE Solution ..........8
5.3.1. Preserving the IGP Hierarchy Concept ................8
5.3.2. Preserving Scalability ..............................8
6. Application Scenario.............................................9
7. Detailed Requirements for Inter-Area MPLS TE ...................10
7.1. Inter-Area MPLS TE Operations and Interoperability ........10
7.2. Inter-Area TE-LSP Signaling ...............................10
7.3. Path Optimality ...........................................11
7.4. Inter-Area MPLS-TE Routing ................................11
7.5. Inter-Area MPLS-TE Path Computation .......................12
7.6. Inter-Area Crankback Routing ..............................12
7.7. Support of Diversely-Routed Inter-Area TE LSPs ............13
7.8. Intra/Inter-Area Path Selection Policy ....................13
7.9. Reoptimization of Inter-Area TE LSP .......................13
7.10. Inter-Area LSP Recovery ..................................14
7.10.1. Rerouting of Inter-Area TE LSPs ..................14
7.10.2. Fast Recovery of Inter-Area TE LSP ...............14
7.11. DS-TE support ............................................15
7.12. Hierarchical LSP Support .................................15
7.13. Hard/Soft Preemption .....................................15
7.14. Auto-Discovery of TE Meshes ..............................16
7.15. Inter-Area MPLS TE Fault Management Requirements .........16
7.16. Inter-Area MPLS TE and Routing ...........................16
8. Evaluation criteria ............................................17
8.1. Performances ..............................................17
8.2. Complexity and Risks ......................................17
8.3. Backward Compatibility ....................................17
9. Security Considerations ........................................17
10. Acknowledgements ..............................................17
11. Contributing Authors ..........................................18
12. Normative References ..........................................19
13. Informative References ........................................19
1. Introduction
The set of MPLS Traffic Engineering components, defined in [RSVP-TE],
[OSPF-TE], and [ISIS-TE], which supports the requirements defined in
[TE-REQ], is used today by many network operators to achieve major
Traffic Engineering objectives defined in [TE-OVW]. These objectives
include:
- Aggregated Traffic measurement
- Optimization of network resources utilization
- Support for services requiring end-to-end QoS guarantees
- Fast recovery against link/node/Shared Risk Link Group (SRLG)
failures
Furthermore, the applicability of MPLS to traffic engineering in IP
networks is discussed in [TE-APP].
The set of MPLS Traffic Engineering mechanisms, to date, has been
limited to use within a single Interior Gateway Protocol (IGP) area.
This document discusses the requirements for an inter-area MPLS
Traffic Engineering mechanism that may be used to achieve the same
set of objectives across multiple IGP areas.
Basically, it would be useful to extend MPLS TE capabilities across
IGP areas to support inter-area resources optimization, to provide
strict QoS guarantees between two edge routers located within
distinct areas, and to protect inter-area traffic against Area Border
Router (ABR) failures.
First, this document addresses current uses of MPLS Traffic
Engineering within a single IGP area. Then, it discusses a set of
functional requirements that a solution must or should satisfy in
order to support inter-area MPLS Traffic Engineering. Because the
scope of requirements will vary between operators, some requirements
will be mandatory (MUST), whereas others will be optional (SHOULD).
Finally, a set of evaluation criteria for any solution meeting these
requirements is given.
2. Conventions Used in This Document
The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT",
"SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this
document are to be interpreted as described in [RFC2119].
3. Terminology
LSR: Label Switching Router
LSP: Label Switched Path
TE LSP: Traffic Engineering Label Switched Path
Inter-area TE LSP: TE LSP whose head-end LSR and tail-end LSR do not
reside within the same IGP area or whose head-end
LSR and tail-end LSR are both in the same IGP area
although the TE-LSP transiting path is across
different IGP areas.
IGP area: OSPF area or IS-IS level.
ABR: Area Border Router, a router used to connect two
IGP areas (ABR in OSPF, or L1/L2 router in IS-IS).
CSPF: Constraint-based Shortest Path First.
SRLG: Shared Risk Link Group.
4. Current Intra-Area Uses of MPLS Traffic Engineering
This section addresses architecture, capabilities, and uses of MPLS
TE within a single IGP area. It first summarizes the current MPLS-TE
architecture, then addresses various MPLS-TE capabilities, and
finally lists various approaches to integrate MPLS TE into routing.
This section is intended to help define the requirements for MPLS-TE
extensions across multiple IGP areas.
4.1. Intra-Area MPLS Traffic Engineering Architecture
The MPLS-TE control plane allows establishing explicitly routed MPLS
LSPs whose paths follow a set of TE constraints. It is used to
achieve major TE objectives such as resource usage optimization, QoS
guarantee and fast failure recovery. It consists of three main
components:
- The routing component, responsible for the discovery of the TE
topology. This is ensured thanks to extensions of link state IGP:
[ISIS-TE], [OSPF-TE].
- The path computation component, responsible for the placement of
the LSP. It is performed on the head-end LSR thanks to a CSPF
algorithm, which takes TE topology and LSP constraints as input.
- The signaling component, responsible for the establishment of the
LSP (explicit routing, label distribution, and resources
reservation) along the computed path. This is ensured thanks to
RSVP-TE [RSVP-TE].
4.2. Intra-Area MPLS Traffic Engineering Applications
4.2.1. Intra-Area Resource Optimization
MPLS TE can be used within an area to redirect paths of aggregated
flows away from over-utilized resources within a network. In a small
scale, this may be done by explicitly configuring a path to be used
between two routers. On a grander scale, a mesh of LSPs can be
established between central points in a network. LSPs paths can be
defined statically in configuration or arrived at by an algorithm
that determines the shortest path given administrative constraints
such as bandwidth. In this way, MPLS TE allows for greater control
over how traffic demands are routed over a network topology and
utilize a network’s resources.
Note also that TE LSPs allow measuring traffic matrix in a simple and
scalable manner. The aggregated traffic rate between two LSRs is
easily measured by accounting of traffic sent onto a TE LSP
provisioned between the two LSRs in question.
4.2.2. Intra-Area QoS Guarantees
The DiffServ IETF working group has defined a set of mechanisms
described in [DIFF-ARCH], [DIFF-AF], and [DIFF-EF] or [MPLS-DIFF],
that can be activated at the edge of or over a DiffServ domain to
contribute to the enforcement of a QoS policy (or set of policies),
which can be expressed in terms of maximum one-way transit delay,
inter-packet delay variation, loss rate, etc. Many Operators have
some or full deployment of DiffServ implementations in their networks
today, either across the entire network or at least at its edge.
In situations where strict QoS bounds are required, admission control
inside the backbone of a network is in some cases required in
addition to current DiffServ mechanisms. When the propagation delay
can be bounded, the performance targets, such as maximum one-way
transit delay, may be guaranteed by providing bandwidth guarantees
along the DiffServ-enabled path.
MPLS TE can be simply used with DiffServ: in that case, it only
ensures aggregate QoS guarantees for the whole traffic. It can also
be more intimately combined with DiffServ to perform per-class of
service admission control and resource reservation. This requires
extensions to MPLS TE called DiffServ-Aware TE, which are defined in
[DSTE-PROTO]. DS-TE allows ensuring strict end-to-end QoS
guarantees. For instance, an EF DS-TE LSP may be provisioned between
voice gateways within the same area to ensure strict QoS to VoIP
traffic.
MPLS TE allows computing intra-area shortest paths, which satisfy
various constraints, including bandwidth. For the sake of
illustration, if the IGP metrics reflects the propagation delay, it
allows finding a minimum propagation delay path, which satisfies
various constraints, such as bandwidth.
4.2.3. Fast Recovery within an IGP Area
As quality-sensitive applications are deployed, one of the key
requirements is to provide fast recovery mechanisms, allowing traffic
recovery to be guaranteed on the order of tens of msecs, in case of
network element failure. Note that this cannot be achieved by
relying only on classical IGP rerouting.
Various recovery mechanisms can be used to protect traffic carried
onto TE LSPs. They are defined in [MPLS-RECOV]. Protection
mechanisms are based on the provisioning of backup LSPs that are used
to recover traffic in case of failure of protected LSPs. Among those
protection mechanisms, local protection (also called Fast Reroute) is
intended to achieve sub-50ms recovery in case of link/node/SRLG
failure along the LSP path [FAST-REROUTE]. Fast Reroute is currently
used by many operators to protect sensitive traffic inside an IGP
area.
[FAST-REROUTE] defines two modes for backup LSPs. The first, called
one-to-one backup, consists of setting up one detour LSP per
protected LSP and per element to protect. The second, called
facility backup, consists of setting up one or several bypass LSPs to
protect a given facility (link or node). In case of failure, all
protected LSPs are nested into the bypass LSPs (benefiting from the
MPLS label stacking property).
4.3. Intra-Area MPLS TE and Routing
There are several possibilities for directing traffic into intra-area
TE LSPs:
1) Static routing to the LSP destination address or any other
addresses.
2) IGP routes beyond the LSP destination, from an IGP SPF perspective
(IGP shortcuts).
3) BGP routes announced by a BGP peer (or an MP-BGP peer) that is
reachable through the TE LSP by means of a single static route to
the corresponding BGP next-hop address (option 1) or by means of
IGP shortcuts (option 2). This is often called BGP recursive
routing.
4) The LSP can be advertised as a link into the IGP to become part of
IGP database for all nodes, and thus can be taken into account
during SPF for all nodes. Note that, even if similar in concept,
this is different from the notion of Forwarding-Adjacency, as
defined in [LSP-HIER]. Forwarding-Adjacency is when the LSP is
advertised as a TE-link into the IGP-TE to become part of the TE
database and taken into account in CSPF.
5. Problem Statement, Requirements, and Objectives of Inter-Area
MPLS TE
5.1. Inter-Area Traffic Engineering Problem Statement
As described in Section 4, MPLS TE is deployed today by many
operators to optimize network bandwidth usage, to provide strict QoS
guarantees, and to ensure sub-50ms recovery in case of link/node/SRLG
failure.
However, MPLS-TE mechanisms are currently limited to a single IGP
area. The limitation comes more from the Routing and Path
computation components than from the signaling component. This is
basically because the hierarchy limits topology visibility of head-
end LSRs to their IGP area, and consequently head-end LSRs can no
longer run a CSPF algorithm to compute the shortest constrained path
to the tail-end, as CSPF requires the whole topology to compute an
end-to-end shortest constrained path.
Several operators have multi-area networks, and many operators that
are still using a single IGP area may have to migrate to a multi-area
environment, as their network grows and single area scalability
limits are approached.
Thus, those operators may require inter-area traffic engineering to:
- Perform inter-area resource optimization.
- Provide inter-area QoS guarantees for traffic between edge nodes
located in different areas.
- Provide fast recovery across areas, to protect inter-area traffic
in case of link or node failure, including ABR node failures.
For instance, an operator running a multi-area IGP may have voice
gateways located in different areas. Such VoIP transport requires
inter-area QoS guarantees and inter-area fast protection.
One possible approach for inter-area traffic engineering could
consist of deploying MPLS TE on a per-area basis, but such an
approach has several limitations:
- Traffic aggregation at the ABR levels implies some constraints that
do not lead to efficient traffic engineering. Actually, this per-
area TE approach might lead to sub-optimal resource utilization, by
optimizing resources independently in each area. What many
operators want is to optimize their resources as a whole; in other
words, as if there was only one area (flat network).
- This does not allow computing an inter-area constrained shortest
path and thus does not ensure end-to-end QoS guarantees across
areas.
- Inter-area traffic cannot be protected with local protection
mechanisms such as [FAST-REROUTE] in case of ABR failure.
Therefore, existing MPLS TE mechanisms have to be enhanced to support
inter-area TE LSPs.
5.2. Overview of Requirements for Inter-Area MPLS TE
For the reasons mentioned above, it is highly desired to extend the
current set of MPLS-TE mechanisms across multiple IGP areas in order
to support the intra-area applications described in Section 4 across
areas.
The solution MUST allow setting up inter-area TE LSPs; i.e., LSPs
whose path crosses at least two IGP areas.
Inter-area MPLS-TE extensions are highly desired in order to provide:
- Inter-area resources optimization.
- Strict inter-area QoS guarantees.
- Fast recovery across areas, particularly to protect inter-area
traffic against ABR failures.
It may be desired to compute inter-area shortest paths that satisfy
some bandwidth constraints or any other constraints, as is currently
possible within a single IGP area. For the sake of illustration, if
the IGP metrics reflects the propagation delay, it may be necessary
to be able to find the optimal (shortest) path satisfying some
constraints (e.g., bandwidth) across multiple IGP areas. Such a path
would be the inter-area path offering the minimal propagation delay.
Thus, the solution SHOULD provide the ability to compute inter-area
shortest paths satisfying a set of constraints (i.e., bandwidth).
5.3. Key Objectives for an Inter-Area MPLS-TE Solution
Any solution for inter-area MPLS TE should be designed with
preserving IGP hierarchy concept, and preserving routing and
signaling scalability as key objectives.
5.3.1. Preserving the IGP Hierarchy Concept
The absence of a full link-state topology database makes the
computation of an end-to-end optimal path by the head-end LSR not
possible without further signaling and routing extensions. There are
several reasons that network operators choose to break up their
network into different areas. These often include scalability and
containment of routing information. The latter can help isolate most
of a network from receiving and processing updates that are of no
consequence to its routing decisions. Containment of routing
information MUST not be compromised to allow inter-area traffic
engineering. Information propagation for path-selection MUST
continue to be localized. In other words, the solution MUST entirely
preserve the concept of IGP hierarchy.
5.3.2. Preserving Scalability
Achieving the requirements listed in this document MUST be performed
while preserving the IGP scalability, which is of the utmost
importance. The hierarchy preservation objective addressed in the
above section is actually an element to preserve IGP scalability.
The solution also MUST not increase IGP load unreasonably, which
could compromise IGP scalability. In particular, a solution
satisfying those requirements MUST not require the IGP to carry some
unreasonable amount of extra information and MUST not unreasonably
increase the IGP flooding frequency.
Likewise, the solution MUST also preserve scalability of RSVP-TE
([RSVP-TE]).
Additionally, the base specification of MPLS TE is architecturally
structured and relatively devoid of excessive state propagation in
terms of routing or signaling. Its strength in extensibility can
also be seen as an Achilles heel, as there is no real limit to what
is possible with extensions. It is paramount to maintain
architectural vision and discretion when adapting it for use for
inter-area MPLS TE. Additional information carried within an area or
propagated outside of an area (via routing or signaling) should be
neither excessive, patchwork, nor non-relevant.
Particularly, as mentioned in Section 5.2, it may be desired for some
inter-area TE LSP carrying highly sensitive traffic to compute a
shortest inter-area path, satisfying a set of constraints such as
bandwidth. This may require an additional routing mechanism, as base
CSPF at head-end can no longer be used due to the lack of topology
and resource information. Such a routing mechanism MUST not
compromise the scalability of the overall system.
6. Application Scenario
---area1--------area0------area2--
------R1-ABR1-R2-------ABR3-------
| \ | / | |
| R0 \ | / | R4 |
| R5 \ |/ | |
---------ABR2----------ABR4-------
- ABR1, ABR2: Area0-Area1 ABRs
- ABR3, ABR4: Area0-Area2 ABRs
- R0, R1, R5: LSRs in area 1
- R2: an LSR in area 0
- R4: an LSR in area 2
Although the terminology and examples provided in this document make
use of the OSPF terminology, this document equally applies to IS-IS.
Typically, an inter-area TE LSP will be set up between R0 and R4,
where both LSRs belong to different IGP areas. Note that the
solution MUST support the capability to protect such an inter-area TE
LSP from the failure on any Link/SRLG/Node within any area and the
failure of any traversed ABR. For instance, if the TE LSP R0->R4
goes through R1->ABR1->R2, then it can be protected against ABR1
failure, thanks to a backup LSP (detour or bypass) that may follow
the alternate path R1->ABR2->R2.
For instance, R0 and R4 may be two voice gateways located in distinct
areas. An inter-area DS-TE LSP with class-type EF is set up from R1
to R4 to route VoIP traffic classified as EF. Per-class inter-area
constraint-based routing allows the DS-TE LSP to be routed over a
path that will ensure strict QoS guarantees for VoIP traffic.
In another application, R0 and R4 may be two pseudo wire gateways
residing in different areas. An inter-area LSP may be set up to
carry pseudo wires.
In some cases, it might also be possible to have an inter-area TE LSP
from R0 to R5 transiting via the backbone area (or any other levels
with IS-IS). There may be cases where there are no longer enough
resources on any intra area path R0-to-R5, and where there is a
feasible inter-area path through the backbone area.
7. Detailed Requirements for Inter-Area MPLS TE
7.1. Inter-Area MPLS TE Operations and Interoperability
The inter-area MPLS TE solution MUST be consistent with requirements
discussed in [TE-REQ], and the derived solution MUST interoperate
seamlessly with current intra-area MPLS TE mechanisms and inherit its
capability sets from [RSVP-TE].
The proposed solution MUST allow provisioning at the head-end with
end-to-end RSVP signaling (potentially with loose paths) traversing
across the interconnected ABRs, without further provisioning required
along the transit path.
7.2. Inter-Area TE-LSP Signaling
The solution MUST allow for the signaling of inter-area TE LSPs,
using RSVP-TE.
In addition to the signaling of classical TE constraints (bandwidth,
admin-groups), the proposed solution MUST allow the head-end LSR to
specify a set of LSRs explicitly, including ABRs, by means of strict
or loose hops for the inter-area TE LSP.
In addition, the proposed solution SHOULD also provide the ability to
specify and signal certain resources to be explicitly excluded in the
inter-area TE-LSP path establishment.
7.3. Path Optimality
In the context of this requirement document, an optimal path is
defined as the shortest path across multiple areas, taking into
account either the IGP or TE metric [METRIC]. In other words, such a
path is the path that would have been computed by making use of some
CSPF algorithm in the absence of multiple IGP areas.
As mentioned in Section 5.2, the solution SHOULD provide the
capability to compute an optimal path dynamically, satisfying a set